Re: %gs:0 thread pseudoregister in oskit-mach

[Roland McGrath]

> No, the assert happens using the DS segment. Hmm, I guess something
> more is wrong here. It looks like the segment isn't protecting us here
> and the segmentation fault just happens because of there is nothing
> mapped at 0xc0000000 and it gets a page fault. If the segment would
> protect us we should get a general protection exception, right?

Yes, I think that's what should happen.  However, I think the segment
limits are all off by one.  That is, try accessing 0xc0000004.  If that is
the only problem, then using VM_MAX_ADDRESS-VM_MIN_ADDRESS-1 in the
fill_ldt_descriptor calls should fix it.